這項研究使用一種以統計概念開發的 IC 測試模型，並以測試良率與測試品質 做為測試結果的評斷標準。此模型以常態分佈為基礎，納入多項表達製造能力與測 試能力的參數，並輔以測試防護帶與重複測試的概念，衍伸出一套嶄新的測試機制， 其中，重複測試包含了重複測試方案與多重測試方案。 對於重複測試，不管使用哪一種測試方案都能帶來測試良率與品質的提升，但 在使用前述的數學模型時，無法快速的得出適合的測試參數，故我們對其進行分析 並獲得遞迴表示式，在此數學模型的基礎下，進一步推演出迭代法(iteration method) 與二分法(bisection method)的自動規劃測試防護帶的測試系統(auto-guardbanding)， 其能在固定測試品質與次數的條件下，迅速得出適當的測試規格。 對於重複測試中的兩種測試方案，其包括的測試與製造的隨機與不穩定特性， 使我們無法直觀的評判出兩者的優劣，但在大量數據的比較下得到在相同測試品 質的要求下，重複測試方案會有較好的測試良率。;This study uses a statistically developed IC test model with test yield and test quality as the criteria for evaluating test results. The model is based on the normative distribution and incorporates a number of parameters that express manufacturability and testability, and is supplemented by the concept of test shields and re-testing, which proposes a new testing mechanism. Among them, the re-testing includes the re-testing scheme and the multi-testing scheme. However, when using the aforementioned mathematical model, it is not possible to quickly derive suitable test parameters, so we analyze it and obtain a recursive expression. Based on this mathematical model, the iteration method and the bisection method are further deduced from the automatic-guardbanding test system, which can quickly derive the corresponding test specifications with a fixed test quality and frequency. The random and unstable characteristics of the test and manufacturing of the two test solutions in the retest make it impossible for us to evaluate the advantages and disadvantages of the two solutions. But under the comparison of a large amount of data, the repeat test method will have a better test yield under the same test quality requirements.
